Patchwork opkg: Update svn 625 -> 633 and fix preinst issues

login
register
mail settings
Submitter Richard Purdie
Date Dec. 15, 2011, 9:08 p.m.
Message ID <1323983329.4568.63.camel@ted>
Download mbox | patch
Permalink /patch/17041/
State Accepted
Commit 2feba313c991170747381c7cf821a45c2cd04632
Headers show

Comments

Richard Purdie - Dec. 15, 2011, 9:08 p.m.
There is a major issue with opkg images at the moment as preinst
functions are not being executed before their dependencies are installed
and this is leading to corruption of images containing avahi/dbus in
particular.

There are various changes in upstream opkg in the last 8 revisions which
make changes in this area but sadly these aren't enough to get things
working for us. I've updated to the latest svn revision with this patch
since it makes sense to pull in those changes first and then supplement
them with the attached patches.

There is a full description of the patches in the patch headers but in
summary they:

a) Ensure preinst functions execute with their dependencies installed.
   This is a pretty invasive change as it changes the package install
   ordering in general.
b) Ensure opkg sets $D, not $PKG_ROOT which we don't use
c) Change opkg to allow execution of postinstall functions which fail
   resulting in execution on the target device as rootfs_ipk.bbclass
   currently does manually.

The remaining changes interface this with the rest of the OE build
infrastructure, adding in the option to tell opkg to run the preinst and
postinst functions, ensure the correct environment is present for the
postinst scripts and removing the now unneeded rootfs_ipk class code
which opkg now does itself.

[YOCTO #1711]
    
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Martin Jansa - Dec. 17, 2011, 1:16 a.m.
On Thu, Dec 15, 2011 at 09:08:49PM +0000, Richard Purdie wrote:
> There is a major issue with opkg images at the moment as preinst
> functions are not being executed before their dependencies are installed
> and this is leading to corruption of images containing avahi/dbus in
> particular.
> 
> There are various changes in upstream opkg in the last 8 revisions which
> make changes in this area but sadly these aren't enough to get things
> working for us. I've updated to the latest svn revision with this patch
> since it makes sense to pull in those changes first and then supplement
> them with the attached patches.
> 
> There is a full description of the patches in the patch headers but in
> summary they:
> 
> a) Ensure preinst functions execute with their dependencies installed.
>    This is a pretty invasive change as it changes the package install
>    ordering in general.
> b) Ensure opkg sets $D, not $PKG_ROOT which we don't use
> c) Change opkg to allow execution of postinstall functions which fail
>    resulting in execution on the target device as rootfs_ipk.bbclass
>    currently does manually.
> 
> The remaining changes interface this with the rest of the OE build
> infrastructure, adding in the option to tell opkg to run the preinst and
> postinst functions, ensure the correct environment is present for the
> postinst scripts and removing the now unneeded rootfs_ipk class code
> which opkg now does itself.
> 
> [YOCTO #1711]

Hi,

today I got image build failing with this and it seems like some kind of
circular dependency or something.

In fsogsmd_git.bb we have:
PACKAGES =+ "${PN}-connman ${PN}-connman-dev ${PN}-connman-dbg"
RDEPENDS_${PN} += "${PN}-connman"

and there is also fsogsmd-connman dependency on fsogsmd (because of shlibs),
so it's really circular dependency between runtime packages and new opkg cannot
deal with it.

Then in log.do_rootfs
Installing fsogsmd (1:0.5.0+gitr20+96698077b6ccd5fc1222e0b2b9fc56d087c6c528-r6.7) to root...
Downloading file:/OE/shr-core/tmp-eglibc/deploy/ipk/qemuarm/fsogsmd_0.5.0+gitr20+96698077b6ccd5fc1222e0b2b9fc56d087c6c528-r6.7_qemuarm.ipk.
Installing fsogsmd-connman (1:0.5.0+gitr20+96698077b6ccd5fc1222e0b2b9fc56d087c6c528-r6.7) to root...
Downloading file:/OE/shr-core/tmp-eglibc/deploy/ipk/qemuarm/fsogsmd-connman_0.5.0+gitr20+96698077b6ccd5fc1222e0b2b9fc56d087c6c528-r6.7_qemuarm.ipk.
Installing fsogsmd (1:0.5.0+gitr20+96698077b6ccd5fc1222e0b2b9fc56d087c6c528-r6.7) to root...
Installing fsogsmd-connman (1:0.5.0+gitr20+96698077b6ccd5fc1222e0b2b9fc56d087c6c528-r6.7) to root...
and then this last 2 lines 1794/2 times
Installing fsogsmd (1:0.5.0+gitr20+96698077b6ccd5fc1222e0b2b9fc56d087c6c528-r6.7) to root...
Installing fsogsmd-connman (1:0.5.0+gitr20+96698077b6ccd5fc1222e0b2b9fc56d087c6c528-r6.7) to root...
Installing task-core-ssh-openssh (1.0-r0) to root...
Downloading file:/OE/shr-core/tmp-eglibc/deploy/ipk/armv5te/task-core-ssh-openssh_1.0-r0_armv5te.ipk.
...
and in the end:
ERROR: Function 'do_rootfs' failed (see /OE/shr-core/tmp-eglibc/work/qemuarm-oe-linux-gnueabi/shr-lite-image/2.0-r20/temp/log.do_rootfs.20387 for further information)
Collected errors:
 * gz_close: Unzip process killed by signal 11.

 * pkg_get_installed_files: Error extracting file list from /tmp/opkg-T7G08e/fsogsmd-connman_0.5.0+gitr20+96698077b6ccd5fc1222e0b2b9fc56d087c6c528-r6.7_qemuarm.ipk.
 * opkg_install_cmd: Cannot install package task-shr-minimal-apps.

Reverting this patch and rebuilding opkg before running image build helps.

I'll try to upgrade only opkg without your patches first, but I guess it's
because of fix_installorder.patch, can we teach it to break circular dependencies
or should I fix it in metadata? It seem that longer cycles like

fsogsmd-config -> fsogsmd-module-modem-ti-calypso -> fsogsmd -> fsogsmd-config

does work.

Regards,

>     
> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
> 
> diff --git a/meta/classes/package_ipk.bbclass b/meta/classes/package_ipk.bbclass
> index 1633aff..aeabc11 100644
> --- a/meta/classes/package_ipk.bbclass
> +++ b/meta/classes/package_ipk.bbclass
> @@ -133,7 +133,7 @@ package_install_internal_ipk() {
>  
>  	mkdir -p ${target_rootfs}${localstatedir}/lib/opkg/
>  
> -	local ipkg_args="-f ${conffile} -o ${target_rootfs} --force-overwrite"
> +	local ipkg_args="-f ${conffile} -o ${target_rootfs} --force-overwrite --force_postinstall"
>  
>  	opkg-cl ${ipkg_args} update
>  
> diff --git a/meta/classes/rootfs_ipk.bbclass b/meta/classes/rootfs_ipk.bbclass
> index ec01951..b5556fa 100644
> --- a/meta/classes/rootfs_ipk.bbclass
> +++ b/meta/classes/rootfs_ipk.bbclass
> @@ -60,14 +60,14 @@ fakeroot rootfs_ipk_do_rootfs () {
>  	export INSTALL_CONF_IPK="${IPKGCONF_TARGET}"
>  	export INSTALL_PACKAGES_IPK="${PACKAGE_INSTALL}"
>  
> -	package_install_internal_ipk
> -
>  	#post install
>  	export D=${IMAGE_ROOTFS}
>  	export OFFLINE_ROOT=${IMAGE_ROOTFS}
>  	export IPKG_OFFLINE_ROOT=${IMAGE_ROOTFS}
>  	export OPKG_OFFLINE_ROOT=${IPKG_OFFLINE_ROOT}
>  
> +	package_install_internal_ipk
> +
>  	# Distro specific packages should create this
>  	#mkdir -p ${IMAGE_ROOTFS}/etc/opkg/
>  	#grep "^arch" ${IPKGCONF_TARGET} >${IMAGE_ROOTFS}/etc/opkg/arch.conf
> @@ -75,28 +75,8 @@ fakeroot rootfs_ipk_do_rootfs () {
>  	${OPKG_POSTPROCESS_COMMANDS}
>  	${ROOTFS_POSTINSTALL_COMMAND}
>  	
> -	runtime_script_required=0
> -
> -	# Base-passwd needs to run first to install /etc/passwd and friends
> -	if [ -e ${IMAGE_ROOTFS}${opkglibdir}/info/base-passwd.preinst ] ; then
> -		sh ${IMAGE_ROOTFS}${opkglibdir}/info/base-passwd.preinst
> -	fi
> -
> -	for i in ${IMAGE_ROOTFS}${opkglibdir}/info/*.preinst; do
> -		if [ -f $i ] && ! sh $i; then
> -		     	runtime_script_required=1
> -			opkg-cl ${IPKG_ARGS} flag unpacked `basename $i .preinst`
> -		fi
> -	done
> -	for i in ${IMAGE_ROOTFS}${opkglibdir}/info/*.postinst; do
> -		if [ -f $i ] && ! sh $i configure; then
> -		     	runtime_script_required=1
> -			opkg-cl ${IPKG_ARGS} flag unpacked `basename $i .postinst`
> -		fi
> -	done
> -
>  	if ${@base_contains("IMAGE_FEATURES", "read-only-rootfs", "true", "false" ,d)}; then
> -		if [ $runtime_script_required -eq 1 ]; then
> +		if grep Status:.install.ok.unpacked ${IMAGE_ROOTFS}${opkglibdir}status; then
>  			echo "Some packages could not be configured offline and rootfs is read-only."
>  			exit 1
>  		fi
> diff --git a/meta/recipes-devtools/opkg/opkg/fix_installorder.patch b/meta/recipes-devtools/opkg/opkg/fix_installorder.patch
> new file mode 100644
> index 0000000..e782ce7
> --- /dev/null
> +++ b/meta/recipes-devtools/opkg/opkg/fix_installorder.patch
> @@ -0,0 +1,174 @@
> +There is a problem with dependency order when installing packages. The key 
> +problem revolves around the satisfy_dependencies_for() function which is 
> +called from opkg_install_pkg just before the installation (and preinst)
> +happens.
> +
> +The satisfy_dependencies_for() function calls pkg_hash_fetch_unsatisfied_dependencies()
> +which will only return packages which were previously not marked as 
> +*going* to be installed at some point. For the purposes of 
> +opkg_install_pkg() we really need to know which dependencies haven't been 
> +installed yet.
> +
> +This patch adds pkg_hash_fetch_satisfied_dependencies() which returns a 
> +list of package dependencies. We can then directly check the status of 
> +these and ensure any hard dependencies (not suggestions or recommendations)
> +are installed before returning.
> +
> +Consider the situation (where -> means 'depends on'):
> +
> +X -> A,E
> +A -> B,E
> +E -> B
> +B -> C
> +
> +Currently X would install A and E. When installing A the packages B, E 
> +and C would be marked as "to install". When the package B is considered
> +the second time (as a dependency of E rather than A), it would install 
> +straight away even though C was not currently installed, just marked
> +as needing to be installed.
> +
> +The patch changes the behaviour so B can't install until C really is installed.
> +
> +This change is required to run the postinst scripts in the correct order.
> +
> +Upstream-Status: Pending
> +
> +RP 2011/12/15
> +
> +Index: trunk/libopkg/opkg_install.c
> +===================================================================
> +--- trunk.orig/libopkg/opkg_install.c	2011-12-15 15:58:39.000000000 +0000
> ++++ trunk/libopkg/opkg_install.c	2011-12-15 15:58:41.838334788 +0000
> +@@ -76,6 +77,27 @@
> +      }
> + 
> +      if (ndepends <= 0) {
> ++	  pkg_vec_free(depends);      
> ++	  depends = pkg_hash_fetch_satisfied_dependencies(pkg);
> ++
> ++	  for (i = 0; i < depends->len; i++) {
> ++	       dep = depends->pkgs[i];
> ++	       /* The package was uninstalled when we started, but another
> ++	          dep earlier in this loop may have depended on it and pulled
> ++	          it in, so check first. */
> ++	       if ((dep->state_status != SS_INSTALLED) && (dep->state_status != SS_UNPACKED)) {
> ++		    opkg_msg(DEBUG2,"Calling opkg_install_pkg.\n");
> ++		    err = opkg_install_pkg(dep, 0);
> ++		    /* mark this package as having been automatically installed to
> ++		     * satisfy a dependancy */
> ++		    dep->auto_installed = 1;
> ++		    if (err) {
> ++			 pkg_vec_free(depends);
> ++			 return err;
> ++		    }
> ++	       }
> ++	  }
> ++
> + 	  pkg_vec_free(depends);
> + 	  return 0;
> +      }
> +Index: trunk/libopkg/pkg_depends.c
> +===================================================================
> +--- trunk.orig/libopkg/pkg_depends.c	2010-12-22 16:04:43.000000000 +0000
> ++++ trunk/libopkg/pkg_depends.c	2011-12-15 15:58:41.838334788 +0000
> +@@ -259,6 +259,88 @@
> +      return unsatisfied->len;
> + }
> + 
> ++
> ++pkg_vec_t *
> ++pkg_hash_fetch_satisfied_dependencies(pkg_t * pkg)
> ++{
> ++     pkg_vec_t *satisfiers;
> ++     int i, j, k;
> ++     int count;
> ++     abstract_pkg_t * ab_pkg;
> ++
> ++     satisfiers = pkg_vec_alloc();
> ++
> ++     /*
> ++      * this is a setup to check for redundant/cyclic dependency checks,
> ++      * which are marked at the abstract_pkg level
> ++      */
> ++     if (!(ab_pkg = pkg->parent)) {
> ++	  opkg_msg(ERROR, "Internal error, with pkg %s.\n", pkg->name);
> ++	  return satisfiers;
> ++     }
> ++
> ++     count = pkg->pre_depends_count + pkg->depends_count + pkg->recommends_count + pkg->suggests_count;
> ++     if (!count)
> ++	  return satisfiers;
> ++
> ++     /* foreach dependency */
> ++     for (i = 0; i < count; i++) {
> ++	  compound_depend_t * compound_depend = &pkg->depends[i];
> ++	  depend_t ** possible_satisfiers = compound_depend->possibilities;;
> ++
> ++          if (compound_depend->type == RECOMMEND || compound_depend->type == SUGGEST)
> ++              continue;
> ++
> ++	  if (compound_depend->type == GREEDY_DEPEND) {
> ++	       /* foreach possible satisfier */
> ++	       for (j = 0; j < compound_depend->possibility_count; j++) {
> ++		    /* foreach provided_by, which includes the abstract_pkg itself */
> ++		    abstract_pkg_t *abpkg = possible_satisfiers[j]->pkg;
> ++		    abstract_pkg_vec_t *ab_provider_vec = abpkg->provided_by;
> ++		    int nposs = ab_provider_vec->len;
> ++		    abstract_pkg_t **ab_providers = ab_provider_vec->pkgs;
> ++		    int l;
> ++		    for (l = 0; l < nposs; l++) {
> ++			 pkg_vec_t *test_vec = ab_providers[l]->pkgs;
> ++			 /* if no depends on this one, try the first package that Provides this one */
> ++			 if (!test_vec){   /* no pkg_vec hooked up to the abstract_pkg!  (need another feed?) */
> ++			      continue;
> ++			 }
> ++
> ++			 /* cruise this possiblity's pkg_vec looking for an installed version */
> ++			 for (k = 0; k < test_vec->len; k++) {
> ++			      pkg_t *pkg_scout = test_vec->pkgs[k];
> ++			      /* not installed, and not already known about? */
> ++			      if (pkg_scout->state_want == SW_INSTALL && pkg_scout != pkg)
> ++      			          pkg_vec_insert(satisfiers, pkg_scout);
> ++			 }
> ++		    }
> ++	       }
> ++
> ++	       continue;
> ++	  }
> ++
> ++	  /* foreach possible satisfier, look for installed package  */
> ++	  for (j = 0; j < compound_depend->possibility_count; j++) {
> ++	       /* foreach provided_by, which includes the abstract_pkg itself */
> ++	       depend_t *dependence_to_satisfy = possible_satisfiers[j];
> ++	       abstract_pkg_t *satisfying_apkg = possible_satisfiers[j]->pkg;
> ++	       pkg_t *satisfying_pkg =
> ++		    pkg_hash_fetch_best_installation_candidate(satisfying_apkg,
> ++							       pkg_installed_and_constraint_satisfied,
> ++							       dependence_to_satisfy, 1);
> ++               /* Being that I can't test constraing in pkg_hash, I will test it here */
> ++	       if (satisfying_pkg != NULL && satisfying_pkg != pkg) {
> ++                  if (pkg_constraint_satisfied(satisfying_pkg, dependence_to_satisfy) && satisfying_pkg->state_want == SW_INSTALL)
> ++	              pkg_vec_insert(satisfiers, satisfying_pkg);
> ++               }
> ++
> ++	  }
> ++     }
> ++     return satisfiers;
> ++}
> ++
> ++
> + /*checking for conflicts !in replaces
> +   If a packages conflicts with another but is also replacing it, I should not consider it a
> +   really conflicts
> +Index: trunk/libopkg/pkg_depends.h
> +===================================================================
> +--- trunk.orig/libopkg/pkg_depends.h	2010-12-22 16:04:43.000000000 +0000
> ++++ trunk/libopkg/pkg_depends.h	2011-12-15 15:58:41.838334788 +0000
> +@@ -82,6 +82,7 @@
> + void buildDependedUponBy(pkg_t * pkg, abstract_pkg_t * ab_pkg);
> + int version_constraints_satisfied(depend_t * depends, pkg_t * pkg);
> + int pkg_hash_fetch_unsatisfied_dependencies(pkg_t * pkg, pkg_vec_t *depends, char *** unresolved);
> ++pkg_vec_t * pkg_hash_fetch_satisfied_dependencies(pkg_t * pkg);
> + pkg_vec_t * pkg_hash_fetch_conflicts(pkg_t * pkg);
> + int pkg_dependence_satisfiable(depend_t *depend);
> + int pkg_dependence_satisfied(depend_t *depend);
> diff --git a/meta/recipes-devtools/opkg/opkg/offline_postinstall.patch b/meta/recipes-devtools/opkg/opkg/offline_postinstall.patch
> new file mode 100644
> index 0000000..b197f6b
> --- /dev/null
> +++ b/meta/recipes-devtools/opkg/opkg/offline_postinstall.patch
> @@ -0,0 +1,57 @@
> +When we have an offline root and have specified force-postinstall, 
> +attempt to run the postinstall but if it fails, just leave it in the
> +status file as neeing to run. We can issue a NOTICE this is happened
> +but supress errors. This means the OE class doesn't have to do any
> +further post processing of the postinstalls itself.
> +
> +Upstream-Status: Pending
> +
> +RP 2011/12/15
> +
> +
> +Index: trunk/libopkg/pkg.c
> +===================================================================
> +--- trunk.orig/libopkg/pkg.c	2011-12-15 15:58:39.000000000 +0000
> ++++ trunk/libopkg/pkg.c	2011-12-15 20:04:50.109992736 +0000
> +@@ -1297,8 +1297,9 @@
> +      free(cmd);
> + 
> +      if (err) {
> +-	  opkg_msg(ERROR, "package \"%s\" %s script returned status %d.\n", 
> +-               pkg->name, script, err);
> ++          if (!conf->offline_root)
> ++	       opkg_msg(ERROR, "package \"%s\" %s script returned status %d.\n", 
> ++                    pkg->name, script, err);
> + 	  return err;
> +      }
> + 
> +Index: trunk/libopkg/opkg_cmd.c
> +===================================================================
> +--- trunk.orig/libopkg/opkg_cmd.c	2011-12-15 19:49:25.826014150 +0000
> ++++ trunk/libopkg/opkg_cmd.c	2011-12-15 19:50:52.346012148 +0000
> +@@ -453,7 +453,8 @@
> + 		    pkg->state_flag &= ~SF_PREFER;
> + 		    opkg_state_changed++;
> + 	       } else {
> +-		    err = -1;
> ++                    if (!conf->offline_root)
> ++		         err = -1;
> + 	       }
> + 	  }
> +      }
> +Index: trunk/libopkg/opkg_configure.c
> +===================================================================
> +--- trunk.orig/libopkg/opkg_configure.c	2011-12-15 19:50:11.586013081 +0000
> ++++ trunk/libopkg/opkg_configure.c	2011-12-15 19:52:15.082010347 +0000
> +@@ -35,7 +35,10 @@
> + 
> +     err = pkg_run_script(pkg, "postinst", "configure");
> +     if (err) {
> +-	opkg_msg(ERROR, "%s.postinst returned %d.\n", pkg->name, err);
> ++        if (!conf->offline_root)
> ++	     opkg_msg(ERROR, "%s.postinst returned %d.\n", pkg->name, err);
> ++        else
> ++	     opkg_msg(NOTICE, "%s.postinst returned %d, marking as unpacked only, configuration required on target.\n", pkg->name, err);
> + 	return err;
> +     }
> + 
> diff --git a/meta/recipes-devtools/opkg/opkg/offlineroot_varname.patch b/meta/recipes-devtools/opkg/opkg/offlineroot_varname.patch
> new file mode 100644
> index 0000000..a67f389
> --- /dev/null
> +++ b/meta/recipes-devtools/opkg/opkg/offlineroot_varname.patch
> @@ -0,0 +1,20 @@
> +OE uses D as the offline install directory in its scripts, not PKG_ROOT.
> +
> +Upstream-Status: Inappropriate [OE specific usage]
> +
> +RP 2011/12/15
> +
> +Index: trunk/libopkg/pkg.c
> +===================================================================
> +--- trunk.orig/libopkg/pkg.c	2011-12-15 15:58:39.000000000 +0000
> ++++ trunk/libopkg/pkg.c	2011-12-15 20:04:50.109992736 +0000
> +@@ -1280,7 +1280,7 @@
> + 
> +      opkg_msg(INFO, "Running script %s.\n", path);
> + 
> +-     setenv("PKG_ROOT",
> ++     setenv("D",
> + 	    pkg->dest ? pkg->dest->root_dir : conf->default_dest->root_dir, 1);
> + 
> +      if (! file_exists(path)) {
> +
> diff --git a/meta/recipes-devtools/opkg/opkg_svn.bb b/meta/recipes-devtools/opkg/opkg_svn.bb
> index acb21f2..2645d52 100644
> --- a/meta/recipes-devtools/opkg/opkg_svn.bb
> +++ b/meta/recipes-devtools/opkg/opkg_svn.bb
> @@ -12,13 +12,16 @@ RREPLACES_${PN} = "opkg-nogpg"
>  SRC_URI = "svn://opkg.googlecode.com/svn;module=trunk;proto=http \
>             file://add_vercmp.patch \
>             file://add_uname_support.patch \
> +           file://fix_installorder.patch \
> +           file://offline_postinstall.patch\
> +           file://offlineroot_varname.patch \
>  "
>  
>  S = "${WORKDIR}/trunk"
>  
> -SRCREV = "625"
> +SRCREV = "633"
>  PV = "0.1.8+svnr${SRCPV}"
> -PR = "r4"
> +PR = "r2"
>  
>  PACKAGES =+ "libopkg${PKGSUFFIX}-dev libopkg${PKGSUFFIX} update-alternatives-cworth${PKGSUFFIX}"
>  
> 
> 
> _______________________________________________
> Openembedded-core mailing list
> Openembedded-core@lists.openembedded.org
> http://lists.linuxtogo.org/cgi-bin/mailman/listinfo/openembedded-core
Richard Purdie - Dec. 17, 2011, 1:32 a.m.
On Sat, 2011-12-17 at 02:16 +0100, Martin Jansa wrote:
> On Thu, Dec 15, 2011 at 09:08:49PM +0000, Richard Purdie wrote:
> > There is a major issue with opkg images at the moment as preinst
> > functions are not being executed before their dependencies are installed
> > and this is leading to corruption of images containing avahi/dbus in
> > particular.
> > 
> > There are various changes in upstream opkg in the last 8 revisions which
> > make changes in this area but sadly these aren't enough to get things
> > working for us. I've updated to the latest svn revision with this patch
> > since it makes sense to pull in those changes first and then supplement
> > them with the attached patches.
> > 
> > There is a full description of the patches in the patch headers but in
> > summary they:
> > 
> > a) Ensure preinst functions execute with their dependencies installed.
> >    This is a pretty invasive change as it changes the package install
> >    ordering in general.
> > b) Ensure opkg sets $D, not $PKG_ROOT which we don't use
> > c) Change opkg to allow execution of postinstall functions which fail
> >    resulting in execution on the target device as rootfs_ipk.bbclass
> >    currently does manually.
> > 
> > The remaining changes interface this with the rest of the OE build
> > infrastructure, adding in the option to tell opkg to run the preinst and
> > postinst functions, ensure the correct environment is present for the
> > postinst scripts and removing the now unneeded rootfs_ipk class code
> > which opkg now does itself.
> > 
> > [YOCTO #1711]
> 
> Hi,
> 
> today I got image build failing with this and it seems like some kind of
> circular dependency or something.
> 
> In fsogsmd_git.bb we have:
> PACKAGES =+ "${PN}-connman ${PN}-connman-dev ${PN}-connman-dbg"
> RDEPENDS_${PN} += "${PN}-connman"
> 
> and there is also fsogsmd-connman dependency on fsogsmd (because of shlibs),
> so it's really circular dependency between runtime packages and new opkg cannot
> deal with it.
> 
> Then in log.do_rootfs
> Installing fsogsmd (1:0.5.0+gitr20+96698077b6ccd5fc1222e0b2b9fc56d087c6c528-r6.7) to root...
> Downloading file:/OE/shr-core/tmp-eglibc/deploy/ipk/qemuarm/fsogsmd_0.5.0+gitr20+96698077b6ccd5fc1222e0b2b9fc56d087c6c528-r6.7_qemuarm.ipk.
> Installing fsogsmd-connman (1:0.5.0+gitr20+96698077b6ccd5fc1222e0b2b9fc56d087c6c528-r6.7) to root...
> Downloading file:/OE/shr-core/tmp-eglibc/deploy/ipk/qemuarm/fsogsmd-connman_0.5.0+gitr20+96698077b6ccd5fc1222e0b2b9fc56d087c6c528-r6.7_qemuarm.ipk.
> Installing fsogsmd (1:0.5.0+gitr20+96698077b6ccd5fc1222e0b2b9fc56d087c6c528-r6.7) to root...
> Installing fsogsmd-connman (1:0.5.0+gitr20+96698077b6ccd5fc1222e0b2b9fc56d087c6c528-r6.7) to root...
> and then this last 2 lines 1794/2 times
> Installing fsogsmd (1:0.5.0+gitr20+96698077b6ccd5fc1222e0b2b9fc56d087c6c528-r6.7) to root...
> Installing fsogsmd-connman (1:0.5.0+gitr20+96698077b6ccd5fc1222e0b2b9fc56d087c6c528-r6.7) to root...
> Installing task-core-ssh-openssh (1.0-r0) to root...
> Downloading file:/OE/shr-core/tmp-eglibc/deploy/ipk/armv5te/task-core-ssh-openssh_1.0-r0_armv5te.ipk.
> ...
> and in the end:
> ERROR: Function 'do_rootfs' failed (see /OE/shr-core/tmp-eglibc/work/qemuarm-oe-linux-gnueabi/shr-lite-image/2.0-r20/temp/log.do_rootfs.20387 for further information)
> Collected errors:
>  * gz_close: Unzip process killed by signal 11.
> 
>  * pkg_get_installed_files: Error extracting file list from /tmp/opkg-T7G08e/fsogsmd-connman_0.5.0+gitr20+96698077b6ccd5fc1222e0b2b9fc56d087c6c528-r6.7_qemuarm.ipk.
>  * opkg_install_cmd: Cannot install package task-shr-minimal-apps.
> 
> Reverting this patch and rebuilding opkg before running image build helps.
> 
> I'll try to upgrade only opkg without your patches first, but I guess it's
> because of fix_installorder.patch, can we teach it to break circular dependencies
> or should I fix it in metadata? It seem that longer cycles like
> 
> fsogsmd-config -> fsogsmd-module-modem-ti-calypso -> fsogsmd -> fsogsmd-config
> 
> does work.

This is likely as a result of the installorder patch :/. Sadly if we
want the postinstalls to be installed in order we do need that patch.

The question is whether circular depends are something opkg should
support or not? What's debian's behaviour in that regard?

Cheers,

Richard
Martin Jansa - Dec. 17, 2011, 9:20 a.m.
On Sat, Dec 17, 2011 at 01:32:28AM +0000, Richard Purdie wrote:
> On Sat, 2011-12-17 at 02:16 +0100, Martin Jansa wrote:
> > On Thu, Dec 15, 2011 at 09:08:49PM +0000, Richard Purdie wrote:
> > > There is a major issue with opkg images at the moment as preinst
> > > functions are not being executed before their dependencies are installed
> > > and this is leading to corruption of images containing avahi/dbus in
> > > particular.
> > > 
> > > There are various changes in upstream opkg in the last 8 revisions which
> > > make changes in this area but sadly these aren't enough to get things
> > > working for us. I've updated to the latest svn revision with this patch
> > > since it makes sense to pull in those changes first and then supplement
> > > them with the attached patches.
> > > 
> > > There is a full description of the patches in the patch headers but in
> > > summary they:
> > > 
> > > a) Ensure preinst functions execute with their dependencies installed.
> > >    This is a pretty invasive change as it changes the package install
> > >    ordering in general.
> > > b) Ensure opkg sets $D, not $PKG_ROOT which we don't use
> > > c) Change opkg to allow execution of postinstall functions which fail
> > >    resulting in execution on the target device as rootfs_ipk.bbclass
> > >    currently does manually.
> > > 
> > > The remaining changes interface this with the rest of the OE build
> > > infrastructure, adding in the option to tell opkg to run the preinst and
> > > postinst functions, ensure the correct environment is present for the
> > > postinst scripts and removing the now unneeded rootfs_ipk class code
> > > which opkg now does itself.
> > > 
> > > [YOCTO #1711]
> > 
> > Hi,
> > 
> > today I got image build failing with this and it seems like some kind of
> > circular dependency or something.
> > 
> > In fsogsmd_git.bb we have:
> > PACKAGES =+ "${PN}-connman ${PN}-connman-dev ${PN}-connman-dbg"
> > RDEPENDS_${PN} += "${PN}-connman"
> > 
> > and there is also fsogsmd-connman dependency on fsogsmd (because of shlibs),
> > so it's really circular dependency between runtime packages and new opkg cannot
> > deal with it.
> > 
> > Then in log.do_rootfs
> > Installing fsogsmd (1:0.5.0+gitr20+96698077b6ccd5fc1222e0b2b9fc56d087c6c528-r6.7) to root...
> > Downloading file:/OE/shr-core/tmp-eglibc/deploy/ipk/qemuarm/fsogsmd_0.5.0+gitr20+96698077b6ccd5fc1222e0b2b9fc56d087c6c528-r6.7_qemuarm.ipk.
> > Installing fsogsmd-connman (1:0.5.0+gitr20+96698077b6ccd5fc1222e0b2b9fc56d087c6c528-r6.7) to root...
> > Downloading file:/OE/shr-core/tmp-eglibc/deploy/ipk/qemuarm/fsogsmd-connman_0.5.0+gitr20+96698077b6ccd5fc1222e0b2b9fc56d087c6c528-r6.7_qemuarm.ipk.
> > Installing fsogsmd (1:0.5.0+gitr20+96698077b6ccd5fc1222e0b2b9fc56d087c6c528-r6.7) to root...
> > Installing fsogsmd-connman (1:0.5.0+gitr20+96698077b6ccd5fc1222e0b2b9fc56d087c6c528-r6.7) to root...
> > and then this last 2 lines 1794/2 times
> > Installing fsogsmd (1:0.5.0+gitr20+96698077b6ccd5fc1222e0b2b9fc56d087c6c528-r6.7) to root...
> > Installing fsogsmd-connman (1:0.5.0+gitr20+96698077b6ccd5fc1222e0b2b9fc56d087c6c528-r6.7) to root...
> > Installing task-core-ssh-openssh (1.0-r0) to root...
> > Downloading file:/OE/shr-core/tmp-eglibc/deploy/ipk/armv5te/task-core-ssh-openssh_1.0-r0_armv5te.ipk.
> > ...
> > and in the end:
> > ERROR: Function 'do_rootfs' failed (see /OE/shr-core/tmp-eglibc/work/qemuarm-oe-linux-gnueabi/shr-lite-image/2.0-r20/temp/log.do_rootfs.20387 for further information)
> > Collected errors:
> >  * gz_close: Unzip process killed by signal 11.
> > 
> >  * pkg_get_installed_files: Error extracting file list from /tmp/opkg-T7G08e/fsogsmd-connman_0.5.0+gitr20+96698077b6ccd5fc1222e0b2b9fc56d087c6c528-r6.7_qemuarm.ipk.
> >  * opkg_install_cmd: Cannot install package task-shr-minimal-apps.
> > 
> > Reverting this patch and rebuilding opkg before running image build helps.
> > 
> > I'll try to upgrade only opkg without your patches first, but I guess it's
> > because of fix_installorder.patch, can we teach it to break circular dependencies
> > or should I fix it in metadata? It seem that longer cycles like
> > 
> > fsogsmd-config -> fsogsmd-module-modem-ti-calypso -> fsogsmd -> fsogsmd-config
> > 
> > does work.
> 
> This is likely as a result of the installorder patch :/. Sadly if we
> want the postinstalls to be installed in order we do need that patch.

I've tried to build image with SRCREV=633 and it built image fine
then I tried to apply your patch but without installorder.patch to
confirm that the issue with circular depending packages is really cause
by that only and then try to fix installorder.patch to work even with
such packages, but I got many postinst errors :(. 

From your description I understood that postinst execution should be
almost the same only more logic is moved from rootfs_ipk.bbclass to opkg
itself.

| Collected errors:
|  * pkg_run_script: package "pam-plugin-unix" postinst script returned status 1.
|  * opkg_configure: pam-plugin-unix.postinst returned 1.
|  * pkg_run_script: package "pango-module-basic-x" postinst script returned status 1.
|  * opkg_configure: pango-module-basic-x.postinst returned 1.
|  * pkg_run_script: package "pango-module-basic-fc" postinst script returned status 1.
|  * opkg_configure: pango-module-basic-fc.postinst returned 1.
|  * pkg_run_script: package "update-modules" postinst script returned status 1.
|  * opkg_configure: update-modules.postinst returned 1.
|  * pkg_run_script: package "gdk-pixbuf-loader-png" postinst script returned status 1.
|  * opkg_configure: gdk-pixbuf-loader-png.postinst returned 1.
|  * pkg_run_script: package "gdk-pixbuf-loader-jpeg" postinst script returned status 1.
|  * opkg_configure: gdk-pixbuf-loader-jpeg.postinst returned 1.
|  * pkg_run_script: package "liberation-fonts" postinst script returned status 1.
|  * opkg_configure: liberation-fonts.postinst returned 1.
|  * pkg_run_script: package "gdk-pixbuf-loader-xpm" postinst script returned status 1.
|  * opkg_configure: gdk-pixbuf-loader-xpm.postinst returned 1.
|  * pkg_run_script: package "gdk-pixbuf-loader-gif" postinst script returned status 1.
|  * opkg_configure: gdk-pixbuf-loader-gif.postinst returned 1.
|  * pkg_run_script: package "ppp" postinst script returned status 1.
|  * opkg_configure: ppp.postinst returned 1.
|  * pkg_run_script: package "ttf-dejavu-common" postinst script returned status 127.
|  * opkg_configure: ttf-dejavu-common.postinst returned 127.
|  * pkg_run_script: package "ttf-dejavu-sans" postinst script returned status 127.
|  * opkg_configure: ttf-dejavu-sans.postinst returned 127.
|  * pkg_run_script: package "ffalarms" postinst script returned status 127.
|  * opkg_configure: ffalarms.postinst returned 127.
|  * pkg_run_script: package "rsyslog" postinst script returned status 1.
|  * opkg_configure: rsyslog.postinst returned 1.
|  * pkg_run_script: package "ttf-dejavu-sans-mono" postinst script returned status 127.
|  * opkg_configure: ttf-dejavu-sans-mono.postinst returned 127.
|  * pkg_run_script: package "matchbox-keyboard-im" postinst script returned status 1.
|  * opkg_configure: matchbox-keyboard-im.postinst returned 1.
|  * pkg_run_script: package "hicolor-icon-theme" postinst script returned status 1.
|  * opkg_configure: hicolor-icon-theme.postinst returned 1.
|  * pkg_run_script: packaERROR: Function 'do_rootfs' failed (see /OE/shr-core/tmp-eglibc/work/qemuarm-oe-linux-gnueabi/shr-l
ite-image/2.0-r20/temp/log.do_rootfs.10063 for further information)
| ge "ttf-liberation-mono" postinst script returned status 127.
|  * opkg_configure: ttf-liberation-mono.postinst returned 127.
|  * pkg_run_script: package "gtk-immodule-xim" postinst script returned status 1.
|  * opkg_configure: gtk-immodule-xim.postinst returned 1.
|  * pkg_run_script: package "ca-certificates" postinst script returned status 1.
|  * opkg_configure: ca-certificates.postinst returned 1.
NOTE: package shr-lite-image-2.0-r20: task do_rootfs: Failed
ERROR: Task 9 (/OE/shr-core/meta-smartphone/meta-shr/recipes-shr/images/shr-lite-image.bb, do_rootfs) failed with exit code '
1'

> The question is whether circular depends are something opkg should
> support or not? What's debian's behaviour in that regard?

Well, it did work before and from quick search it looks that such circular dependencies are 
quite common in our metadata. So I agree that this patch is really needed to fix dbus owner,
but in case there is circular dependency we shouldn't be so strict about order of installs.

Regards,
Richard Purdie - Dec. 17, 2011, 10:22 a.m.
On Sat, 2011-12-17 at 10:20 +0100, Martin Jansa wrote:
> On Sat, Dec 17, 2011 at 01:32:28AM +0000, Richard Purdie wrote:
> > On Sat, 2011-12-17 at 02:16 +0100, Martin Jansa wrote:
> > > On Thu, Dec 15, 2011 at 09:08:49PM +0000, Richard Purdie wrote:
> > > > There is a major issue with opkg images at the moment as preinst
> > > > functions are not being executed before their dependencies are installed
> > > > and this is leading to corruption of images containing avahi/dbus in
> > > > particular.
> > > > 
> > > > There are various changes in upstream opkg in the last 8 revisions which
> > > > make changes in this area but sadly these aren't enough to get things
> > > > working for us. I've updated to the latest svn revision with this patch
> > > > since it makes sense to pull in those changes first and then supplement
> > > > them with the attached patches.
> > > > 
> > > > There is a full description of the patches in the patch headers but in
> > > > summary they:
> > > > 
> > > > a) Ensure preinst functions execute with their dependencies installed.
> > > >    This is a pretty invasive change as it changes the package install
> > > >    ordering in general.
> > > > b) Ensure opkg sets $D, not $PKG_ROOT which we don't use
> > > > c) Change opkg to allow execution of postinstall functions which fail
> > > >    resulting in execution on the target device as rootfs_ipk.bbclass
> > > >    currently does manually.
> > > > 
> > > > The remaining changes interface this with the rest of the OE build
> > > > infrastructure, adding in the option to tell opkg to run the preinst and
> > > > postinst functions, ensure the correct environment is present for the
> > > > postinst scripts and removing the now unneeded rootfs_ipk class code
> > > > which opkg now does itself.
> > > > 
> > > > [YOCTO #1711]
> > > 
> > > Hi,
> > > 
> > > today I got image build failing with this and it seems like some kind of
> > > circular dependency or something.
> > > 
> > > In fsogsmd_git.bb we have:
> > > PACKAGES =+ "${PN}-connman ${PN}-connman-dev ${PN}-connman-dbg"
> > > RDEPENDS_${PN} += "${PN}-connman"
> > > 
> > > and there is also fsogsmd-connman dependency on fsogsmd (because of shlibs),
> > > so it's really circular dependency between runtime packages and new opkg cannot
> > > deal with it.
> > > 
> > > Then in log.do_rootfs
> > > Installing fsogsmd (1:0.5.0+gitr20+96698077b6ccd5fc1222e0b2b9fc56d087c6c528-r6.7) to root...
> > > Downloading file:/OE/shr-core/tmp-eglibc/deploy/ipk/qemuarm/fsogsmd_0.5.0+gitr20+96698077b6ccd5fc1222e0b2b9fc56d087c6c528-r6.7_qemuarm.ipk.
> > > Installing fsogsmd-connman (1:0.5.0+gitr20+96698077b6ccd5fc1222e0b2b9fc56d087c6c528-r6.7) to root...
> > > Downloading file:/OE/shr-core/tmp-eglibc/deploy/ipk/qemuarm/fsogsmd-connman_0.5.0+gitr20+96698077b6ccd5fc1222e0b2b9fc56d087c6c528-r6.7_qemuarm.ipk.
> > > Installing fsogsmd (1:0.5.0+gitr20+96698077b6ccd5fc1222e0b2b9fc56d087c6c528-r6.7) to root...
> > > Installing fsogsmd-connman (1:0.5.0+gitr20+96698077b6ccd5fc1222e0b2b9fc56d087c6c528-r6.7) to root...
> > > and then this last 2 lines 1794/2 times
> > > Installing fsogsmd (1:0.5.0+gitr20+96698077b6ccd5fc1222e0b2b9fc56d087c6c528-r6.7) to root...
> > > Installing fsogsmd-connman (1:0.5.0+gitr20+96698077b6ccd5fc1222e0b2b9fc56d087c6c528-r6.7) to root...
> > > Installing task-core-ssh-openssh (1.0-r0) to root...
> > > Downloading file:/OE/shr-core/tmp-eglibc/deploy/ipk/armv5te/task-core-ssh-openssh_1.0-r0_armv5te.ipk.
> > > ...
> > > and in the end:
> > > ERROR: Function 'do_rootfs' failed (see /OE/shr-core/tmp-eglibc/work/qemuarm-oe-linux-gnueabi/shr-lite-image/2.0-r20/temp/log.do_rootfs.20387 for further information)
> > > Collected errors:
> > >  * gz_close: Unzip process killed by signal 11.
> > > 
> > >  * pkg_get_installed_files: Error extracting file list from /tmp/opkg-T7G08e/fsogsmd-connman_0.5.0+gitr20+96698077b6ccd5fc1222e0b2b9fc56d087c6c528-r6.7_qemuarm.ipk.
> > >  * opkg_install_cmd: Cannot install package task-shr-minimal-apps.
> > > 
> > > Reverting this patch and rebuilding opkg before running image build helps.
> > > 
> > > I'll try to upgrade only opkg without your patches first, but I guess it's
> > > because of fix_installorder.patch, can we teach it to break circular dependencies
> > > or should I fix it in metadata? It seem that longer cycles like
> > > 
> > > fsogsmd-config -> fsogsmd-module-modem-ti-calypso -> fsogsmd -> fsogsmd-config
> > > 
> > > does work.
> > 
> > This is likely as a result of the installorder patch :/. Sadly if we
> > want the postinstalls to be installed in order we do need that patch.
> 
> I've tried to build image with SRCREV=633 and it built image fine
> then I tried to apply your patch but without installorder.patch to
> confirm that the issue with circular depending packages is really cause
> by that only and then try to fix installorder.patch to work even with
> such packages, but I got many postinst errors :(. 
> 
> From your description I understood that postinst execution should be
> almost the same only more logic is moved from rootfs_ipk.bbclass to opkg
> itself.
> 
> | Collected errors:
> |  * pkg_run_script: package "pam-plugin-unix" postinst script returned status 1.
> |  * opkg_configure: pam-plugin-unix.postinst returned 1.
> |  * pkg_run_script: package "pango-module-basic-x" postinst script returned status 1.
> |  * opkg_configure: pango-module-basic-x.postinst returned 1.
> |  * pkg_run_script: package "pango-module-basic-fc" postinst script returned status 1.
> |  * opkg_configure: pango-module-basic-fc.postinst returned 1.
> |  * pkg_run_script: package "update-modules" postinst script returned status 1.
> |  * opkg_configure: update-modules.postinst returned 1.
> |  * pkg_run_script: package "gdk-pixbuf-loader-png" postinst script returned status 1.
> |  * opkg_configure: gdk-pixbuf-loader-png.postinst returned 1.
> |  * pkg_run_script: package "gdk-pixbuf-loader-jpeg" postinst script returned status 1.
> |  * opkg_configure: gdk-pixbuf-loader-jpeg.postinst returned 1.
> |  * pkg_run_script: package "liberation-fonts" postinst script returned status 1.
> |  * opkg_configure: liberation-fonts.postinst returned 1.
> |  * pkg_run_script: package "gdk-pixbuf-loader-xpm" postinst script returned status 1.
> |  * opkg_configure: gdk-pixbuf-loader-xpm.postinst returned 1.
> |  * pkg_run_script: package "gdk-pixbuf-loader-gif" postinst script returned status 1.
> |  * opkg_configure: gdk-pixbuf-loader-gif.postinst returned 1.
> |  * pkg_run_script: package "ppp" postinst script returned status 1.
> |  * opkg_configure: ppp.postinst returned 1.
> |  * pkg_run_script: package "ttf-dejavu-common" postinst script returned status 127.
> |  * opkg_configure: ttf-dejavu-common.postinst returned 127.
> |  * pkg_run_script: package "ttf-dejavu-sans" postinst script returned status 127.
> |  * opkg_configure: ttf-dejavu-sans.postinst returned 127.
> |  * pkg_run_script: package "ffalarms" postinst script returned status 127.
> |  * opkg_configure: ffalarms.postinst returned 127.
> |  * pkg_run_script: package "rsyslog" postinst script returned status 1.
> |  * opkg_configure: rsyslog.postinst returned 1.
> |  * pkg_run_script: package "ttf-dejavu-sans-mono" postinst script returned status 127.
> |  * opkg_configure: ttf-dejavu-sans-mono.postinst returned 127.
> |  * pkg_run_script: package "matchbox-keyboard-im" postinst script returned status 1.
> |  * opkg_configure: matchbox-keyboard-im.postinst returned 1.
> |  * pkg_run_script: package "hicolor-icon-theme" postinst script returned status 1.
> |  * opkg_configure: hicolor-icon-theme.postinst returned 1.
> |  * pkg_run_script: packaERROR: Function 'do_rootfs' failed (see /OE/shr-core/tmp-eglibc/work/qemuarm-oe-linux-gnueabi/shr-l
> ite-image/2.0-r20/temp/log.do_rootfs.10063 for further information)
> | ge "ttf-liberation-mono" postinst script returned status 127.
> |  * opkg_configure: ttf-liberation-mono.postinst returned 127.
> |  * pkg_run_script: package "gtk-immodule-xim" postinst script returned status 1.
> |  * opkg_configure: gtk-immodule-xim.postinst returned 1.
> |  * pkg_run_script: package "ca-certificates" postinst script returned status 1.
> |  * opkg_configure: ca-certificates.postinst returned 1.
> NOTE: package shr-lite-image-2.0-r20: task do_rootfs: Failed
> ERROR: Task 9 (/OE/shr-core/meta-smartphone/meta-shr/recipes-shr/images/shr-lite-image.bb, do_rootfs) failed with exit code '
> 1'

This looks like you don't have the offline_postinstall patch applied?

opkg revision 633 will not build correct metadata without the patches
included.

> > The question is whether circular depends are something opkg should
> > support or not? What's debian's behaviour in that regard?
> 
> Well, it did work before and from quick search it looks that such circular dependencies are 
> quite common in our metadata.

Do you have some examples? I couldn't find any in OE-Core?

>  So I agree that this patch is really needed to fix dbus owner,
> but in case there is circular dependency we shouldn't be so strict about order of installs.

So how you you handle A depends on B which depends on A and be strict
about dependency order?

Cheers,

Richard
Martin Jansa - Dec. 17, 2011, 10:34 a.m.
On Sat, Dec 17, 2011 at 10:22:01AM +0000, Richard Purdie wrote:
> On Sat, 2011-12-17 at 10:20 +0100, Martin Jansa wrote:
> > On Sat, Dec 17, 2011 at 01:32:28AM +0000, Richard Purdie wrote:
> > > On Sat, 2011-12-17 at 02:16 +0100, Martin Jansa wrote:
> > > > On Thu, Dec 15, 2011 at 09:08:49PM +0000, Richard Purdie wrote:
> > > > > There is a major issue with opkg images at the moment as preinst
> > > > > functions are not being executed before their dependencies are installed
> > > > > and this is leading to corruption of images containing avahi/dbus in
> > > > > particular.
> > > > > 
> > > > > There are various changes in upstream opkg in the last 8 revisions which
> > > > > make changes in this area but sadly these aren't enough to get things
> > > > > working for us. I've updated to the latest svn revision with this patch
> > > > > since it makes sense to pull in those changes first and then supplement
> > > > > them with the attached patches.
> > > > > 
> > > > > There is a full description of the patches in the patch headers but in
> > > > > summary they:
> > > > > 
> > > > > a) Ensure preinst functions execute with their dependencies installed.
> > > > >    This is a pretty invasive change as it changes the package install
> > > > >    ordering in general.
> > > > > b) Ensure opkg sets $D, not $PKG_ROOT which we don't use
> > > > > c) Change opkg to allow execution of postinstall functions which fail
> > > > >    resulting in execution on the target device as rootfs_ipk.bbclass
> > > > >    currently does manually.
> > > > > 
> > > > > The remaining changes interface this with the rest of the OE build
> > > > > infrastructure, adding in the option to tell opkg to run the preinst and
> > > > > postinst functions, ensure the correct environment is present for the
> > > > > postinst scripts and removing the now unneeded rootfs_ipk class code
> > > > > which opkg now does itself.
> > > > > 
> > > > > [YOCTO #1711]
> > > > 
> > > > Hi,
> > > > 
> > > > today I got image build failing with this and it seems like some kind of
> > > > circular dependency or something.
> > > > 
> > > > In fsogsmd_git.bb we have:
> > > > PACKAGES =+ "${PN}-connman ${PN}-connman-dev ${PN}-connman-dbg"
> > > > RDEPENDS_${PN} += "${PN}-connman"
> > > > 
> > > > and there is also fsogsmd-connman dependency on fsogsmd (because of shlibs),
> > > > so it's really circular dependency between runtime packages and new opkg cannot
> > > > deal with it.
> > > > 
> > > > Then in log.do_rootfs
> > > > Installing fsogsmd (1:0.5.0+gitr20+96698077b6ccd5fc1222e0b2b9fc56d087c6c528-r6.7) to root...
> > > > Downloading file:/OE/shr-core/tmp-eglibc/deploy/ipk/qemuarm/fsogsmd_0.5.0+gitr20+96698077b6ccd5fc1222e0b2b9fc56d087c6c528-r6.7_qemuarm.ipk.
> > > > Installing fsogsmd-connman (1:0.5.0+gitr20+96698077b6ccd5fc1222e0b2b9fc56d087c6c528-r6.7) to root...
> > > > Downloading file:/OE/shr-core/tmp-eglibc/deploy/ipk/qemuarm/fsogsmd-connman_0.5.0+gitr20+96698077b6ccd5fc1222e0b2b9fc56d087c6c528-r6.7_qemuarm.ipk.
> > > > Installing fsogsmd (1:0.5.0+gitr20+96698077b6ccd5fc1222e0b2b9fc56d087c6c528-r6.7) to root...
> > > > Installing fsogsmd-connman (1:0.5.0+gitr20+96698077b6ccd5fc1222e0b2b9fc56d087c6c528-r6.7) to root...
> > > > and then this last 2 lines 1794/2 times
> > > > Installing fsogsmd (1:0.5.0+gitr20+96698077b6ccd5fc1222e0b2b9fc56d087c6c528-r6.7) to root...
> > > > Installing fsogsmd-connman (1:0.5.0+gitr20+96698077b6ccd5fc1222e0b2b9fc56d087c6c528-r6.7) to root...
> > > > Installing task-core-ssh-openssh (1.0-r0) to root...
> > > > Downloading file:/OE/shr-core/tmp-eglibc/deploy/ipk/armv5te/task-core-ssh-openssh_1.0-r0_armv5te.ipk.
> > > > ...
> > > > and in the end:
> > > > ERROR: Function 'do_rootfs' failed (see /OE/shr-core/tmp-eglibc/work/qemuarm-oe-linux-gnueabi/shr-lite-image/2.0-r20/temp/log.do_rootfs.20387 for further information)
> > > > Collected errors:
> > > >  * gz_close: Unzip process killed by signal 11.
> > > > 
> > > >  * pkg_get_installed_files: Error extracting file list from /tmp/opkg-T7G08e/fsogsmd-connman_0.5.0+gitr20+96698077b6ccd5fc1222e0b2b9fc56d087c6c528-r6.7_qemuarm.ipk.
> > > >  * opkg_install_cmd: Cannot install package task-shr-minimal-apps.
> > > > 
> > > > Reverting this patch and rebuilding opkg before running image build helps.
> > > > 
> > > > I'll try to upgrade only opkg without your patches first, but I guess it's
> > > > because of fix_installorder.patch, can we teach it to break circular dependencies
> > > > or should I fix it in metadata? It seem that longer cycles like
> > > > 
> > > > fsogsmd-config -> fsogsmd-module-modem-ti-calypso -> fsogsmd -> fsogsmd-config
> > > > 
> > > > does work.
> > > 
> > > This is likely as a result of the installorder patch :/. Sadly if we
> > > want the postinstalls to be installed in order we do need that patch.
> > 
> > I've tried to build image with SRCREV=633 and it built image fine
> > then I tried to apply your patch but without installorder.patch to
> > confirm that the issue with circular depending packages is really cause
> > by that only and then try to fix installorder.patch to work even with
> > such packages, but I got many postinst errors :(. 
> > 
> > From your description I understood that postinst execution should be
> > almost the same only more logic is moved from rootfs_ipk.bbclass to opkg
> > itself.
> > 
> > | Collected errors:
> > |  * pkg_run_script: package "pam-plugin-unix" postinst script returned status 1.
> > |  * opkg_configure: pam-plugin-unix.postinst returned 1.
> > |  * pkg_run_script: package "pango-module-basic-x" postinst script returned status 1.
> > |  * opkg_configure: pango-module-basic-x.postinst returned 1.
> > |  * pkg_run_script: package "pango-module-basic-fc" postinst script returned status 1.
> > |  * opkg_configure: pango-module-basic-fc.postinst returned 1.
> > |  * pkg_run_script: package "update-modules" postinst script returned status 1.
> > |  * opkg_configure: update-modules.postinst returned 1.
> > |  * pkg_run_script: package "gdk-pixbuf-loader-png" postinst script returned status 1.
> > |  * opkg_configure: gdk-pixbuf-loader-png.postinst returned 1.
> > |  * pkg_run_script: package "gdk-pixbuf-loader-jpeg" postinst script returned status 1.
> > |  * opkg_configure: gdk-pixbuf-loader-jpeg.postinst returned 1.
> > |  * pkg_run_script: package "liberation-fonts" postinst script returned status 1.
> > |  * opkg_configure: liberation-fonts.postinst returned 1.
> > |  * pkg_run_script: package "gdk-pixbuf-loader-xpm" postinst script returned status 1.
> > |  * opkg_configure: gdk-pixbuf-loader-xpm.postinst returned 1.
> > |  * pkg_run_script: package "gdk-pixbuf-loader-gif" postinst script returned status 1.
> > |  * opkg_configure: gdk-pixbuf-loader-gif.postinst returned 1.
> > |  * pkg_run_script: package "ppp" postinst script returned status 1.
> > |  * opkg_configure: ppp.postinst returned 1.
> > |  * pkg_run_script: package "ttf-dejavu-common" postinst script returned status 127.
> > |  * opkg_configure: ttf-dejavu-common.postinst returned 127.
> > |  * pkg_run_script: package "ttf-dejavu-sans" postinst script returned status 127.
> > |  * opkg_configure: ttf-dejavu-sans.postinst returned 127.
> > |  * pkg_run_script: package "ffalarms" postinst script returned status 127.
> > |  * opkg_configure: ffalarms.postinst returned 127.
> > |  * pkg_run_script: package "rsyslog" postinst script returned status 1.
> > |  * opkg_configure: rsyslog.postinst returned 1.
> > |  * pkg_run_script: package "ttf-dejavu-sans-mono" postinst script returned status 127.
> > |  * opkg_configure: ttf-dejavu-sans-mono.postinst returned 127.
> > |  * pkg_run_script: package "matchbox-keyboard-im" postinst script returned status 1.
> > |  * opkg_configure: matchbox-keyboard-im.postinst returned 1.
> > |  * pkg_run_script: package "hicolor-icon-theme" postinst script returned status 1.
> > |  * opkg_configure: hicolor-icon-theme.postinst returned 1.
> > |  * pkg_run_script: packaERROR: Function 'do_rootfs' failed (see /OE/shr-core/tmp-eglibc/work/qemuarm-oe-linux-gnueabi/shr-l
> > ite-image/2.0-r20/temp/log.do_rootfs.10063 for further information)
> > | ge "ttf-liberation-mono" postinst script returned status 127.
> > |  * opkg_configure: ttf-liberation-mono.postinst returned 127.
> > |  * pkg_run_script: package "gtk-immodule-xim" postinst script returned status 1.
> > |  * opkg_configure: gtk-immodule-xim.postinst returned 1.
> > |  * pkg_run_script: package "ca-certificates" postinst script returned status 1.
> > |  * opkg_configure: ca-certificates.postinst returned 1.
> > NOTE: package shr-lite-image-2.0-r20: task do_rootfs: Failed
> > ERROR: Task 9 (/OE/shr-core/meta-smartphone/meta-shr/recipes-shr/images/shr-lite-image.bb, do_rootfs) failed with exit code '
> > 1'
> 
> This looks like you don't have the offline_postinstall patch applied?

I have

> 
> opkg revision 633 will not build correct metadata without the patches
> included.

Only patch not included in my opkg build was the installorder one.
 
> > > The question is whether circular depends are something opkg should
> > > support or not? What's debian's behaviour in that regard?
> > 
> > Well, it did work before and from quick search it looks that such circular dependencies are 
> > quite common in our metadata.
> 
> Do you have some examples? I couldn't find any in OE-Core?

I'll try.

> >  So I agree that this patch is really needed to fix dbus owner,
> > but in case there is circular dependency we shouldn't be so strict about order of installs.
> 
> So how you you handle A depends on B which depends on A and be strict
> about dependency order?

The problem with dbus/base-passwd can be fixed by this patch, but for
fsogsmd* modules the order is not so important, so I think that when there
is circular dependency detected the order _shouldn't_ be so strict and
maybe warning can be shown, but it shouldn't be fatal.

> Cheers,
> 
> Richard
> 
> 
> _______________________________________________
> Openembedded-core mailing list
> Openembedded-core@lists.openembedded.org
> http://lists.linuxtogo.org/cgi-bin/mailman/listinfo/openembedded-core
Enrico Scholz - Dec. 18, 2011, 10:37 a.m.
Richard Purdie
<richard.purdie-hQyY1W1yCW8ekmWlsbkhG0B+6BGkLq7r@public.gmane.org>
writes:

> This is likely as a result of the installorder patch :/. Sadly if we
> want the postinstalls to be installed in order we do need that patch.
>
> The question is whether circular depends are something opkg should
> support or not?

I think, circular dependencies should (and must) be supported. It is
nearly impossible to avoid them and other package managers (rpm, dpkg)
work fine with them.

Forbidding them would require a large rewrite of oe recipes; e.g. the
autogenerated perl deps cause a circle leading to

| Installing perl-module-extutils-makemaker (5.14.2-r2) to root...
| Installing perl-module-extutils-my (5.14.2-r2) to root...
| Installing perl-module-extutils-mm (5.14.2-r2) to root...
| Installing perl-module-extutils-makemaker (5.14.2-r2) to root...
| Installing perl-module-extutils-my (5.14.2-r2) to root...
| Installing perl-module-extutils-mm (5.14.2-r2) to root...

loops.


> What's debian's behaviour in that regard?

afais, dpkg supports a 'Pre-Depends' field which breaks the dependency
circles.  rpm supports more finegrained 'Requires(preinst/postinst)'
tags too.


Enrico

Patch

diff --git a/meta/classes/package_ipk.bbclass b/meta/classes/package_ipk.bbclass
index 1633aff..aeabc11 100644
--- a/meta/classes/package_ipk.bbclass
+++ b/meta/classes/package_ipk.bbclass
@@ -133,7 +133,7 @@  package_install_internal_ipk() {
 
 	mkdir -p ${target_rootfs}${localstatedir}/lib/opkg/
 
-	local ipkg_args="-f ${conffile} -o ${target_rootfs} --force-overwrite"
+	local ipkg_args="-f ${conffile} -o ${target_rootfs} --force-overwrite --force_postinstall"
 
 	opkg-cl ${ipkg_args} update
 
diff --git a/meta/classes/rootfs_ipk.bbclass b/meta/classes/rootfs_ipk.bbclass
index ec01951..b5556fa 100644
--- a/meta/classes/rootfs_ipk.bbclass
+++ b/meta/classes/rootfs_ipk.bbclass
@@ -60,14 +60,14 @@  fakeroot rootfs_ipk_do_rootfs () {
 	export INSTALL_CONF_IPK="${IPKGCONF_TARGET}"
 	export INSTALL_PACKAGES_IPK="${PACKAGE_INSTALL}"
 
-	package_install_internal_ipk
-
 	#post install
 	export D=${IMAGE_ROOTFS}
 	export OFFLINE_ROOT=${IMAGE_ROOTFS}
 	export IPKG_OFFLINE_ROOT=${IMAGE_ROOTFS}
 	export OPKG_OFFLINE_ROOT=${IPKG_OFFLINE_ROOT}
 
+	package_install_internal_ipk
+
 	# Distro specific packages should create this
 	#mkdir -p ${IMAGE_ROOTFS}/etc/opkg/
 	#grep "^arch" ${IPKGCONF_TARGET} >${IMAGE_ROOTFS}/etc/opkg/arch.conf
@@ -75,28 +75,8 @@  fakeroot rootfs_ipk_do_rootfs () {
 	${OPKG_POSTPROCESS_COMMANDS}
 	${ROOTFS_POSTINSTALL_COMMAND}
 	
-	runtime_script_required=0
-
-	# Base-passwd needs to run first to install /etc/passwd and friends
-	if [ -e ${IMAGE_ROOTFS}${opkglibdir}/info/base-passwd.preinst ] ; then
-		sh ${IMAGE_ROOTFS}${opkglibdir}/info/base-passwd.preinst
-	fi
-
-	for i in ${IMAGE_ROOTFS}${opkglibdir}/info/*.preinst; do
-		if [ -f $i ] && ! sh $i; then
-		     	runtime_script_required=1
-			opkg-cl ${IPKG_ARGS} flag unpacked `basename $i .preinst`
-		fi
-	done
-	for i in ${IMAGE_ROOTFS}${opkglibdir}/info/*.postinst; do
-		if [ -f $i ] && ! sh $i configure; then
-		     	runtime_script_required=1
-			opkg-cl ${IPKG_ARGS} flag unpacked `basename $i .postinst`
-		fi
-	done
-
 	if ${@base_contains("IMAGE_FEATURES", "read-only-rootfs", "true", "false" ,d)}; then
-		if [ $runtime_script_required -eq 1 ]; then
+		if grep Status:.install.ok.unpacked ${IMAGE_ROOTFS}${opkglibdir}status; then
 			echo "Some packages could not be configured offline and rootfs is read-only."
 			exit 1
 		fi
diff --git a/meta/recipes-devtools/opkg/opkg/fix_installorder.patch b/meta/recipes-devtools/opkg/opkg/fix_installorder.patch
new file mode 100644
index 0000000..e782ce7
--- /dev/null
+++ b/meta/recipes-devtools/opkg/opkg/fix_installorder.patch
@@ -0,0 +1,174 @@ 
+There is a problem with dependency order when installing packages. The key 
+problem revolves around the satisfy_dependencies_for() function which is 
+called from opkg_install_pkg just before the installation (and preinst)
+happens.
+
+The satisfy_dependencies_for() function calls pkg_hash_fetch_unsatisfied_dependencies()
+which will only return packages which were previously not marked as 
+*going* to be installed at some point. For the purposes of 
+opkg_install_pkg() we really need to know which dependencies haven't been 
+installed yet.
+
+This patch adds pkg_hash_fetch_satisfied_dependencies() which returns a 
+list of package dependencies. We can then directly check the status of 
+these and ensure any hard dependencies (not suggestions or recommendations)
+are installed before returning.
+
+Consider the situation (where -> means 'depends on'):
+
+X -> A,E
+A -> B,E
+E -> B
+B -> C
+
+Currently X would install A and E. When installing A the packages B, E 
+and C would be marked as "to install". When the package B is considered
+the second time (as a dependency of E rather than A), it would install 
+straight away even though C was not currently installed, just marked
+as needing to be installed.
+
+The patch changes the behaviour so B can't install until C really is installed.
+
+This change is required to run the postinst scripts in the correct order.
+
+Upstream-Status: Pending
+
+RP 2011/12/15
+
+Index: trunk/libopkg/opkg_install.c
+===================================================================
+--- trunk.orig/libopkg/opkg_install.c	2011-12-15 15:58:39.000000000 +0000
++++ trunk/libopkg/opkg_install.c	2011-12-15 15:58:41.838334788 +0000
+@@ -76,6 +77,27 @@
+      }
+ 
+      if (ndepends <= 0) {
++	  pkg_vec_free(depends);      
++	  depends = pkg_hash_fetch_satisfied_dependencies(pkg);
++
++	  for (i = 0; i < depends->len; i++) {
++	       dep = depends->pkgs[i];
++	       /* The package was uninstalled when we started, but another
++	          dep earlier in this loop may have depended on it and pulled
++	          it in, so check first. */
++	       if ((dep->state_status != SS_INSTALLED) && (dep->state_status != SS_UNPACKED)) {
++		    opkg_msg(DEBUG2,"Calling opkg_install_pkg.\n");
++		    err = opkg_install_pkg(dep, 0);
++		    /* mark this package as having been automatically installed to
++		     * satisfy a dependancy */
++		    dep->auto_installed = 1;
++		    if (err) {
++			 pkg_vec_free(depends);
++			 return err;
++		    }
++	       }
++	  }
++
+ 	  pkg_vec_free(depends);
+ 	  return 0;
+      }
+Index: trunk/libopkg/pkg_depends.c
+===================================================================
+--- trunk.orig/libopkg/pkg_depends.c	2010-12-22 16:04:43.000000000 +0000
++++ trunk/libopkg/pkg_depends.c	2011-12-15 15:58:41.838334788 +0000
+@@ -259,6 +259,88 @@
+      return unsatisfied->len;
+ }
+ 
++
++pkg_vec_t *
++pkg_hash_fetch_satisfied_dependencies(pkg_t * pkg)
++{
++     pkg_vec_t *satisfiers;
++     int i, j, k;
++     int count;
++     abstract_pkg_t * ab_pkg;
++
++     satisfiers = pkg_vec_alloc();
++
++     /*
++      * this is a setup to check for redundant/cyclic dependency checks,
++      * which are marked at the abstract_pkg level
++      */
++     if (!(ab_pkg = pkg->parent)) {
++	  opkg_msg(ERROR, "Internal error, with pkg %s.\n", pkg->name);
++	  return satisfiers;
++     }
++
++     count = pkg->pre_depends_count + pkg->depends_count + pkg->recommends_count + pkg->suggests_count;
++     if (!count)
++	  return satisfiers;
++
++     /* foreach dependency */
++     for (i = 0; i < count; i++) {
++	  compound_depend_t * compound_depend = &pkg->depends[i];
++	  depend_t ** possible_satisfiers = compound_depend->possibilities;;
++
++          if (compound_depend->type == RECOMMEND || compound_depend->type == SUGGEST)
++              continue;
++
++	  if (compound_depend->type == GREEDY_DEPEND) {
++	       /* foreach possible satisfier */
++	       for (j = 0; j < compound_depend->possibility_count; j++) {
++		    /* foreach provided_by, which includes the abstract_pkg itself */
++		    abstract_pkg_t *abpkg = possible_satisfiers[j]->pkg;
++		    abstract_pkg_vec_t *ab_provider_vec = abpkg->provided_by;
++		    int nposs = ab_provider_vec->len;
++		    abstract_pkg_t **ab_providers = ab_provider_vec->pkgs;
++		    int l;
++		    for (l = 0; l < nposs; l++) {
++			 pkg_vec_t *test_vec = ab_providers[l]->pkgs;
++			 /* if no depends on this one, try the first package that Provides this one */
++			 if (!test_vec){   /* no pkg_vec hooked up to the abstract_pkg!  (need another feed?) */
++			      continue;
++			 }
++
++			 /* cruise this possiblity's pkg_vec looking for an installed version */
++			 for (k = 0; k < test_vec->len; k++) {
++			      pkg_t *pkg_scout = test_vec->pkgs[k];
++			      /* not installed, and not already known about? */
++			      if (pkg_scout->state_want == SW_INSTALL && pkg_scout != pkg)
++      			          pkg_vec_insert(satisfiers, pkg_scout);
++			 }
++		    }
++	       }
++
++	       continue;
++	  }
++
++	  /* foreach possible satisfier, look for installed package  */
++	  for (j = 0; j < compound_depend->possibility_count; j++) {
++	       /* foreach provided_by, which includes the abstract_pkg itself */
++	       depend_t *dependence_to_satisfy = possible_satisfiers[j];
++	       abstract_pkg_t *satisfying_apkg = possible_satisfiers[j]->pkg;
++	       pkg_t *satisfying_pkg =
++		    pkg_hash_fetch_best_installation_candidate(satisfying_apkg,
++							       pkg_installed_and_constraint_satisfied,
++							       dependence_to_satisfy, 1);
++               /* Being that I can't test constraing in pkg_hash, I will test it here */
++	       if (satisfying_pkg != NULL && satisfying_pkg != pkg) {
++                  if (pkg_constraint_satisfied(satisfying_pkg, dependence_to_satisfy) && satisfying_pkg->state_want == SW_INSTALL)
++	              pkg_vec_insert(satisfiers, satisfying_pkg);
++               }
++
++	  }
++     }
++     return satisfiers;
++}
++
++
+ /*checking for conflicts !in replaces
+   If a packages conflicts with another but is also replacing it, I should not consider it a
+   really conflicts
+Index: trunk/libopkg/pkg_depends.h
+===================================================================
+--- trunk.orig/libopkg/pkg_depends.h	2010-12-22 16:04:43.000000000 +0000
++++ trunk/libopkg/pkg_depends.h	2011-12-15 15:58:41.838334788 +0000
+@@ -82,6 +82,7 @@
+ void buildDependedUponBy(pkg_t * pkg, abstract_pkg_t * ab_pkg);
+ int version_constraints_satisfied(depend_t * depends, pkg_t * pkg);
+ int pkg_hash_fetch_unsatisfied_dependencies(pkg_t * pkg, pkg_vec_t *depends, char *** unresolved);
++pkg_vec_t * pkg_hash_fetch_satisfied_dependencies(pkg_t * pkg);
+ pkg_vec_t * pkg_hash_fetch_conflicts(pkg_t * pkg);
+ int pkg_dependence_satisfiable(depend_t *depend);
+ int pkg_dependence_satisfied(depend_t *depend);
diff --git a/meta/recipes-devtools/opkg/opkg/offline_postinstall.patch b/meta/recipes-devtools/opkg/opkg/offline_postinstall.patch
new file mode 100644
index 0000000..b197f6b
--- /dev/null
+++ b/meta/recipes-devtools/opkg/opkg/offline_postinstall.patch
@@ -0,0 +1,57 @@ 
+When we have an offline root and have specified force-postinstall, 
+attempt to run the postinstall but if it fails, just leave it in the
+status file as neeing to run. We can issue a NOTICE this is happened
+but supress errors. This means the OE class doesn't have to do any
+further post processing of the postinstalls itself.
+
+Upstream-Status: Pending
+
+RP 2011/12/15
+
+
+Index: trunk/libopkg/pkg.c
+===================================================================
+--- trunk.orig/libopkg/pkg.c	2011-12-15 15:58:39.000000000 +0000
++++ trunk/libopkg/pkg.c	2011-12-15 20:04:50.109992736 +0000
+@@ -1297,8 +1297,9 @@
+      free(cmd);
+ 
+      if (err) {
+-	  opkg_msg(ERROR, "package \"%s\" %s script returned status %d.\n", 
+-               pkg->name, script, err);
++          if (!conf->offline_root)
++	       opkg_msg(ERROR, "package \"%s\" %s script returned status %d.\n", 
++                    pkg->name, script, err);
+ 	  return err;
+      }
+ 
+Index: trunk/libopkg/opkg_cmd.c
+===================================================================
+--- trunk.orig/libopkg/opkg_cmd.c	2011-12-15 19:49:25.826014150 +0000
++++ trunk/libopkg/opkg_cmd.c	2011-12-15 19:50:52.346012148 +0000
+@@ -453,7 +453,8 @@
+ 		    pkg->state_flag &= ~SF_PREFER;
+ 		    opkg_state_changed++;
+ 	       } else {
+-		    err = -1;
++                    if (!conf->offline_root)
++		         err = -1;
+ 	       }
+ 	  }
+      }
+Index: trunk/libopkg/opkg_configure.c
+===================================================================
+--- trunk.orig/libopkg/opkg_configure.c	2011-12-15 19:50:11.586013081 +0000
++++ trunk/libopkg/opkg_configure.c	2011-12-15 19:52:15.082010347 +0000
+@@ -35,7 +35,10 @@
+ 
+     err = pkg_run_script(pkg, "postinst", "configure");
+     if (err) {
+-	opkg_msg(ERROR, "%s.postinst returned %d.\n", pkg->name, err);
++        if (!conf->offline_root)
++	     opkg_msg(ERROR, "%s.postinst returned %d.\n", pkg->name, err);
++        else
++	     opkg_msg(NOTICE, "%s.postinst returned %d, marking as unpacked only, configuration required on target.\n", pkg->name, err);
+ 	return err;
+     }
+ 
diff --git a/meta/recipes-devtools/opkg/opkg/offlineroot_varname.patch b/meta/recipes-devtools/opkg/opkg/offlineroot_varname.patch
new file mode 100644
index 0000000..a67f389
--- /dev/null
+++ b/meta/recipes-devtools/opkg/opkg/offlineroot_varname.patch
@@ -0,0 +1,20 @@ 
+OE uses D as the offline install directory in its scripts, not PKG_ROOT.
+
+Upstream-Status: Inappropriate [OE specific usage]
+
+RP 2011/12/15
+
+Index: trunk/libopkg/pkg.c
+===================================================================
+--- trunk.orig/libopkg/pkg.c	2011-12-15 15:58:39.000000000 +0000
++++ trunk/libopkg/pkg.c	2011-12-15 20:04:50.109992736 +0000
+@@ -1280,7 +1280,7 @@
+ 
+      opkg_msg(INFO, "Running script %s.\n", path);
+ 
+-     setenv("PKG_ROOT",
++     setenv("D",
+ 	    pkg->dest ? pkg->dest->root_dir : conf->default_dest->root_dir, 1);
+ 
+      if (! file_exists(path)) {
+
diff --git a/meta/recipes-devtools/opkg/opkg_svn.bb b/meta/recipes-devtools/opkg/opkg_svn.bb
index acb21f2..2645d52 100644
--- a/meta/recipes-devtools/opkg/opkg_svn.bb
+++ b/meta/recipes-devtools/opkg/opkg_svn.bb
@@ -12,13 +12,16 @@  RREPLACES_${PN} = "opkg-nogpg"
 SRC_URI = "svn://opkg.googlecode.com/svn;module=trunk;proto=http \
            file://add_vercmp.patch \
            file://add_uname_support.patch \
+           file://fix_installorder.patch \
+           file://offline_postinstall.patch\
+           file://offlineroot_varname.patch \
 "
 
 S = "${WORKDIR}/trunk"
 
-SRCREV = "625"
+SRCREV = "633"
 PV = "0.1.8+svnr${SRCPV}"
-PR = "r4"
+PR = "r2"
 
 PACKAGES =+ "libopkg${PKGSUFFIX}-dev libopkg${PKGSUFFIX} update-alternatives-cworth${PKGSUFFIX}"